Artificial Intelligence Nanodegree¶
Convolutional Neural Networks¶
Project: Write an Algorithm for a Dog Identification App¶
In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!
Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.
The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.
Why We're Here¶
In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).
In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!
The Road Ahead¶
We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.
- Step 0: Import Datasets
- Step 1: Detect Humans
- Step 2: Detect Dogs
- Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
- Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 6: Write your Algorithm
- Step 7: Test Your Algorithm
Step 0: Import Datasets¶
Import Dog Dataset¶
In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:
train_files,valid_files,test_files- numpy arrays containing file paths to imagestrain_targets,valid_targets,test_targets- numpy arrays containing onehot-encoded classification labelsdog_names- list of string-valued dog breed names for translating labels
from sklearn.datasets import load_files
from keras.utils import np_utils
import numpy as np
from glob import glob
# define function to load train, test, and validation datasets
def load_dataset(path):
data = load_files(path)
dog_files = np.array(data['filenames'])
dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
return dog_files, dog_targets
# load train, test, and validation datasets
train_files, train_targets = load_dataset('dogImages/train')
valid_files, valid_targets = load_dataset('dogImages/valid')
test_files, test_targets = load_dataset('dogImages/test')
# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("dogImages/train/*/"))]
# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))
Import Human Dataset¶
In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.
import random
random.seed(8675309)
# load filenames in shuffled human dataset
human_files = np.array(glob("lfw/*/*"))
random.shuffle(human_files)
# print statistics about the dataset
print('There are %d total human images.' % len(human_files))
Step 1: Detect Humans¶
We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.
In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.
import cv2
import matplotlib.pyplot as plt
%matplotlib inline
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
# load color (BGR) image
img = cv2.imread(human_files[3])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
faces = face_cascade.detectMultiScale(gray)
# print number of faces detected in the image
print('Number of faces detected:', len(faces))
# get bounding box for each detected face
for (x,y,w,h) in faces:
# add bounding box to color image
cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.
In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.
Write a Human Face Detector¶
We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return len(faces) > 0
(IMPLEMENTATION) Assess the Human Face Detector¶
Question 1: Use the code cell below to test the performance of the face_detector function.
- What percentage of the first 100 images in
human_fileshave a detected human face? - What percentage of the first 100 images in
dog_fileshave a detected human face?
Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.
Answer:
The percentage of the first 100 images in
human_fileswhere a human face is detected is 98%, while the percentage of the first 100 images indog_fileswhere a human face is detected is 12%.
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.
## TODO: Test the performance of the face_detector algorithm
## on the images in human_files_short and dog_files_short.
faces_in_humans = 0
for human_file in human_files_short:
if (face_detector(human_file)):
faces_in_humans += 1
print(f"% of human faces found in human pictures: {faces_in_humans}%")
faces_in_dogs = 0
for dog_file in dog_files_short:
if (face_detector(dog_file)):
faces_in_dogs += 1
print(f"% of human faces found in dog pictures: {faces_in_dogs}%")
Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?
Answer:
I don't think this is a reasonable expectation, particularly in a time where face detection algorithms are getting so accurate regardless of the context (eg. Google Photos). On an initial thought I see 2 ways of detecting humans in images without requiring a clear frontal view:
- By combining the frontal face cascade classifier with the profile cascade classifier
- By using a pre-trained deep neural net
The former approach improves human face classification from 98% to 99%. At the same time, the scope of what's classified as a human face is larger so the risk of false positives also increases. This is visible by the increasing amount of human faces detected in dog pictures, from 12% to 14%. I was not very confident on this approach as I inspected the misclassified images and found no empirical evidence that considering profile images would help. Still, it was worth investigating. The experiment is available on the next cell.
As for the later approach, by carefully tuning the confidence parameter, it improves human face classification from 98% to 100%. It also reduces the amount of false positives in dog pictures, from 12% to 2%. The experiment is available on the cell after the next. It uses a pre-trained DNN from OpenCV, linked to in the cell comments.
We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.
# This attempt revolves around adding the profile cascade classifier, but although it can increase correct human
# face classifications, it would also increase dog false positives.
#
# Diplaying the wrongly classified images shows that none are the case of a profile picture.
#
# This approach shows mixed results, improving human face recognition in human pictures from 98% to 99% but
# increasing human face recognition in dog pictures from 12% to 14%.
frontalface_classifier = cv2.CascadeClassifier("haarcascades/haarcascade_frontalface_alt.xml")
profile_classifier = cv2.CascadeClassifier("haarcascades/haarcascade_profileface.xml")
# returns "True" if face is detected in image stored at img_path
def face_detector2(img_path):
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
frontal_faces = frontalface_classifier.detectMultiScale(gray)
profile_faces = profile_classifier.detectMultiScale(gray)
return len(frontal_faces) + len(profile_faces) > 0
max_examples = 5
non_face_indices = []
for i, human_file in enumerate(human_files_short):
if (not face_detector2(human_file)):
non_face_indices.append(i)
print(f"% of faces in human pictures: {100-len(non_face_indices)}%")
plt.figure(figsize=(20,10))
columns = min(len(non_face_indices), max_examples)
for i, index in enumerate(non_face_indices[:max_examples]):
plt.subplot(2, columns, i + 1)
plt.title("Not a face")
plt.axis("off")
plt.imshow(cv2.cvtColor(cv2.imread(human_files_short[index]), cv2.COLOR_BGR2RGB))
plt.show()
face_indices = []
for i, dog_file in enumerate(dog_files_short):
if (face_detector2(dog_file)):
face_indices.append(i)
print(f"% of faces found in dog pictures: {len(face_indices)}%")
plt.figure(figsize=(20,10))
columns = min(len(face_indices), max_examples)
for i, index in enumerate(face_indices[:max_examples]):
plt.subplot(2, columns, i + 1)
plt.title("Face")
plt.axis("off")
plt.imshow(cv2.cvtColor(cv2.imread(dog_files_short[index]), cv2.COLOR_BGR2RGB))
plt.show()
# This attempt is based on https://github.com/opencv/opencv/tree/master/samples/dnn/face_detector,
# using a DNN with a pre-trained model. This model expects 300x300 pictures and two approaches were tried:
# - Stretching the image (ie. losing its original aspect ratio)
# - Adding black borders (ie. maintaining its original aspect ratio)
# Both perform effectively the same so the first one, which does less image manipulation, is used.
# Curiously, the dog image these approaches mismatch is a different one.
#
# This approach shows a significant performance increase, classifying correctly 100% of human pictures
# and 99% of dog pictures.
ResNet10_model = cv2.dnn.readNetFromCaffe("res10/deploy.prototxt.txt", "res10/res10_300x300_ssd_iter_140000.caffemodel")
desired_size = 300
confidence = 0.997
# Best is 100% accuracy for humans, 99% for dogs.
def get_blob_strecthed(img_path):
img = cv2.imread(img_path)
img = cv2.resize(cv2.imread(img_path), (desired_size, desired_size))
return cv2.dnn.blobFromImage(img, 1.0, (desired_size, desired_size), (104.0, 117.0, 124.0))
# Best is 100% accuracy for humans, 99% for dogs.
def get_blob_stable_aspect_ratio_border(img_path):
img = cv2.imread(img_path)
old_size = img.shape[:2]
ratio = float(desired_size) / max(old_size)
new_size = tuple([int(x*ratio) for x in old_size])
img = cv2.resize(img, (new_size[1], new_size[0]))
delta_w = desired_size - new_size[1]
delta_h = desired_size - new_size[0]
top, bottom = delta_h//2, delta_h-(delta_h//2)
left, right = delta_w//2, delta_w-(delta_w//2)
img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=[0, 0, 0])
return cv2.dnn.blobFromImage(img, 1.0, (desired_size, desired_size), (104.0, 117.0, 123.0))
def face_detector3(img_path):
blob = get_blob_strecthed(img_path)
#blob = get_blob_stable_aspect_ratio_border(img_path)
ResNet10_model.setInput(blob)
detections = ResNet10_model.forward()
return len(list(filter(lambda i: detections[0, 0, i, 2] > confidence, range(len(detections))))) > 0
max_examples = 5
non_human_indices = []
for i, human_file in enumerate(human_files_short):
if (not face_detector3(human_file)):
non_human_indices.append(i)
print(f"% of faces found in human pictures: {100-len(non_human_indices)}%")
plt.figure(figsize=(20,10))
columns = min(len(non_human_indices), max_examples)
for i, index in enumerate(non_human_indices[:max_examples]):
plt.subplot(2, columns, i + 1)
plt.title("Not a face")
plt.axis("off")
plt.imshow(cv2.cvtColor(cv2.imread(human_files_short[index]), cv2.COLOR_BGR2RGB))
plt.show()
human_indices = []
for i, dog_file in enumerate(dog_files_short):
if (face_detector3(dog_file)):
human_indices.append(i)
print(f"% of faces found in dog pictures: {len(human_indices)}%")
plt.figure(figsize=(20,10))
columns = min(len(human_indices), max_examples)
for i, index in enumerate(human_indices[:max_examples]):
plt.subplot(2, columns, i + 1)
plt.title("Face")
plt.axis("off")
plt.imshow(cv2.cvtColor(cv2.imread(dog_files_short[index]), cv2.COLOR_BGR2RGB))
plt.show()
Step 2: Detect Dogs¶
In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.
from keras.applications.resnet50 import ResNet50
# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')
Pre-process the Data¶
When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape
$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$
where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.
The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape
$$ (1, 224, 224, 3). $$
The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape
$$ (\text{nb_samples}, 224, 224, 3). $$
Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!
from keras.preprocessing import image
from tqdm import tqdm
def path_to_tensor(img_path):
# loads RGB image as PIL.Image.Image type
img = image.load_img(img_path, target_size=(224, 224))
# convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
x = image.img_to_array(img)
# convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
return np.expand_dims(x, axis=0)
def paths_to_tensor(img_paths):
list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
return np.vstack(list_of_tensors)
Making Predictions with ResNet-50¶
Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.
Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.
By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.
from keras.applications.resnet50 import preprocess_input, decode_predictions
def ResNet50_predict_labels(img_path):
# returns prediction vector for image located at img_path
img = preprocess_input(path_to_tensor(img_path))
return np.argmax(ResNet50_model.predict(img))
Write a Dog Detector¶
While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).
We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
prediction = ResNet50_predict_labels(img_path)
return ((prediction <= 268) & (prediction >= 151))
(IMPLEMENTATION) Assess the Dog Detector¶
Question 3: Use the code cell below to test the performance of your dog_detector function.
- What percentage of the images in
human_files_shorthave a detected dog? - What percentage of the images in
dog_files_shorthave a detected dog?
Answer:
The percentage of images in
human_files_shortwhere a dog is detected is 100%, while the percentage of images indog_files_shortwhere a dog is detected is 3%.
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
dogs_in_dogs = 0
for dog_file in dog_files_short:
if (dog_detector(dog_file)):
dogs_in_dogs += 1
print(f"% of dogs found in dog pictures: {dogs_in_dogs}%")
dogs_in_humans = 0
for human_file in human_files_short:
if (dog_detector(human_file)):
dogs_in_humans += 1
print(f"% of dogs found in human pictures: {dogs_in_humans}%")
Step 3: Create a CNN to Classify Dog Breeds (from Scratch)¶
Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.
Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.
We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.
| Brittany | Welsh Springer Spaniel |
|---|---|
 |
 |
It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).
| Curly-Coated Retriever | American Water Spaniel |
|---|---|
 |
 |
Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.
| Yellow Labrador | Chocolate Labrador | Black Labrador |
|---|---|---|
 |
 |
 |
We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.
Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!
Pre-process the Data¶
We rescale the images by dividing every pixel in every image by 255.
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:
model.summary()
We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:
Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.
Answer:
I've started with a more complex network than the one proposed above, where multiple (3/2) convolutional layers would preceed each max pooling layer. The performance met the requirements, but hitting 1.5% test accuracy right off the bat made it feel like there was a lot of room for improvement.
The first step was trying a simpler architecture, faster to train and evaluate. It was now closer to the proposed architecture, but with more parameters. Adding additional conv / pooling layers seemed to benefit the network up to around 6 pairs, after which there were clear diminishing returns. Similarly, additional parameters seemed to benefit the network up to the values shown below, after which overfitting seemed too extravagant to work with. These changes alone improved performance significantly, to around 9%. Even in this setup there were clear signs of overfitting, with validation loss decreasing and then increasing rapidly. In fact, there were no improvements after 3~5 epochs. This led to the second step.
The second step revolved around adding dropout layers, to help with overfitting. I started with the final layers: at first with the fully connected layer and last pooling layer, then added the second-to-last pooling layer, etc. The reasoning behind this approach is that the initial layers likely hold very simple patterns which are unlikely to be overfit. Contrary to this, the last few layers contain more complex patterns which are less general, more specific to the training set, and easier to overfit. This increase in complexity and potential fitting happens in cescent, so an attempt was made to scale dropouts in a similar fashion. This combined approach brought performance up to around 16%.
The third step was augmenting the training data. This would likely help training and, at the same time, help prevent overfitting. A few combinations were tested and by using width / height shifts, rotation and horizontal flipping, preformance increased to around 23%.
The forth and final step was a review of epochs and the dropout layers. Now that the data was augmented, my intuition was telling me that we could get away with more epochs and less dropout layers. After some fine tunning of these, performance increased to around 34%.
At this point I didn't invest more in this architecture as besides having accomplished the goal (>1%), further changes were showing diminishing returns and significant improvements would likely require a larger training dataset or major architectural changes to the network.
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
model = Sequential()
model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=128, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.05))
model.add(Conv2D(filters=256, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.1))
model.add(Conv2D(filters=512, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.2))
model.add(Conv2D(filters=1024, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.3))
model.add(Flatten())
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(133, activation='softmax'))
model.summary()
Compile the Model¶
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])
(IMPLEMENTATION) Train the Model¶
Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.
You are welcome to augment the training data, but this is not a requirement.
from keras.callbacks import ModelCheckpoint
from keras.preprocessing.image import ImageDataGenerator
augment = True
epochs = 50
batch_size = 32
checkpointer = ModelCheckpoint(
filepath='saved_models/weights.best.from_scratch.hdf5', verbose=1, save_best_only=True)
if augment:
datagen = ImageDataGenerator(
width_shift_range=0.1, height_shift_range=0.1, rotation_range=30, horizontal_flip=True)
datagen.fit(train_tensors)
model.fit_generator(
datagen.flow(train_tensors, train_targets, batch_size=batch_size),
steps_per_epoch=train_tensors.shape[0] // batch_size, epochs=epochs, verbose=1, workers=8,
callbacks=[checkpointer],
validation_data=(valid_tensors, valid_targets), validation_steps=valid_tensors.shape[0] // batch_size)
else:
model.fit(
train_tensors, train_targets, epochs=epochs, batch_size=batch_size, verbose=1, callbacks=[checkpointer],
validation_data=(valid_tensors, valid_targets))
Load the Model with the Best Validation Loss¶
model.load_weights('saved_models/weights.best.from_scratch.hdf5')
Test the Model¶
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]
# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
bottleneck_features = np.load('bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']
Model Architecture¶
The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))
VGG16_model.summary()
Compile the Model¶
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
Train the Model¶
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5',
verbose=1, save_best_only=True)
VGG16_model.fit(train_VGG16, train_targets,
validation_data=(valid_VGG16, valid_targets),
epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)
Load the Model with the Best Validation Loss¶
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')
Test the Model¶
Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]
# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Predict Dog Breed with the Model¶
from extract_bottleneck_features import *
def VGG16_predict_breed(img_path):
# extract bottleneck features
bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = VGG16_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)¶
You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.
In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras:
- VGG-19 bottleneck features
- ResNet-50 bottleneck features
- Inception bottleneck features
- Xception bottleneck features
The files are encoded as such:
Dog{network}Data.npz
where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception. Pick one of the above architectures, download the corresponding bottleneck features, and store the downloaded file in the bottleneck_features/ folder in the repository.
(IMPLEMENTATION) Obtain Bottleneck Features¶
In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:
bottleneck_features = np.load('bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']bottleneck_features = np.load('bottleneck_features/DogXceptionData.npz')
train_Xception = bottleneck_features['train']
valid_Xception = bottleneck_features['valid']
test_Xception = bottleneck_features['test']
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:
<your model's name>.summary()
Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.
Answer:
The very first step was a quick test to see which CNN appeared more promising for the specific problem at hand, by checking their accuracy after training them for a small number of epochs (5). Xception was chosen, although all performed very similarly and within margin of error of one another.
The architecture built on top of the chosen CNN is very simple, similar to the VGG16 example above. The intuition is that these networks are already heavily tuned and rich with information, and that added complexity would likely not yield significant gains.
The initial simplified approach, as outlined above, yielded an accuracy around 85%. All attempts at adding layers yielded similar or worse results. Similarly, using any other pooling layer type yielded similar or worse results. These results match the original intuition, and consequently the simple approach was taken.
As for augmenting the training data, it was attemped, but didn't affect the network positively. Different approaches were taken, with varying hyper-parameters and network complexity, but test accuracy was never improved over using a non-augmented training set. In fact, it was often worse, at around 80%. My reasoning around these results is based on the same initial principle: these networks are already heavily tuned and rich with information, so augmenting the training data might be resulting in additional entropy.
Finally, I believe the current architecture is suitable for the problem. It is based on an established network trained on the ImageNet dataset, and achieves a high accuracy given how many categories exist. At 85% accuracy the network performs quite admirably, given how complex the task is.
from keras.applications.xception import Xception
from keras.models import Model
augment = False
if augment:
ImageNet_Xception_model = Xception(weights='imagenet', include_top=False)
for layer in ImageNet_Xception_model.layers:
layer.trainable = False
Xception_model = GlobalAveragePooling2D()(ImageNet_Xception_model.output)
Xception_model = Dense(133, activation='softmax')(Xception_model)
Xception_model = Model(inputs=ImageNet_Xception_model.input, outputs=Xception_model)
else:
Xception_model = Sequential()
Xception_model.add(GlobalAveragePooling2D(input_shape=train_Xception.shape[1:]))
Xception_model.add(Dense(133, activation='softmax'))
Xception_model.summary()
(IMPLEMENTATION) Compile the Model¶
Xception_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
(IMPLEMENTATION) Train the Model¶
Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.
You are welcome to augment the training data, but this is not a requirement.
epochs = 20
batch_size = 32
checkpointer = ModelCheckpoint(
filepath='saved_models/weights.best.Xception.hdf5', verbose=1, save_best_only=True)
if augment:
datagen = ImageDataGenerator(
width_shift_range=0.1, height_shift_range=0.1, rotation_range=30, horizontal_flip=True)
datagen.fit(train_tensors)
Xception_model.fit_generator(
datagen.flow(train_tensors, train_targets, batch_size=batch_size),
steps_per_epoch=train_tensors.shape[0] // batch_size, epochs=epochs, verbose=1, workers=8,
callbacks=[checkpointer],
validation_data=(valid_tensors, valid_targets), validation_steps=valid_tensors.shape[0] // batch_size)
else:
Xception_model.fit(
train_Xception, train_targets, epochs=epochs, batch_size=batch_size, verbose=1, callbacks=[checkpointer],
validation_data=(valid_Xception, valid_targets))
(IMPLEMENTATION) Load the Model with the Best Validation Loss¶
Xception_model.load_weights('saved_models/weights.best.Xception.hdf5')
(IMPLEMENTATION) Test the Model¶
Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.
tensors = test_tensors if augment else test_Xception
# get index of predicted dog breed for each image in test set
Xception_predictions = [np.argmax(Xception_model.predict(np.expand_dims(tensor, axis=0))) for tensor in tensors]
# report test accuracy
test_accuracy = 100*np.sum(np.array(Xception_predictions)==np.argmax(test_targets, axis=1))/len(Xception_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
(IMPLEMENTATION) Predict Dog Breed with the Model¶
Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.
Similar to the analogous function in Step 5, your function should have three steps:
- Extract the bottleneck features corresponding to the chosen CNN model.
- Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
- Use the
dog_namesarray defined in Step 0 of this notebook to return the corresponding breed.
The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function
extract_{network}
where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.
from extract_bottleneck_features import *
def Xception_predict_breed(img_path):
# extract bottleneck features
bottleneck_feature = extract_Xception(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = Xception_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
Step 6: Write your Algorithm¶
Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,
- if a dog is detected in the image, return the predicted breed.
- if a human is detected in the image, return the resembling dog breed.
- if neither is detected in the image, provide output that indicates an error.
You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.
Some sample output for our algorithm is provided below, but feel free to design your own user experience!
(IMPLEMENTATION) Write your Algorithm¶
def Xception_predict_breed(img_path):
# extract bottleneck features
bottleneck_feature = extract_Xception(path_to_tensor(img_path))
# obtain predicted vector
predicted_vector = Xception_model.predict(bottleneck_feature)
# return dog breed that is predicted by the model
return dog_names[np.argmax(predicted_vector)]
Step 7: Test Your Algorithm¶
In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?
(IMPLEMENTATION) Test Your Algorithm on Sample Images!¶
Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.
Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.
Answer:
On one hand, the output is better than expected. Most images in the
otherfolder (cats, cow and goats) were classified as dogs by Google Photos. In this algorithm, though, only one of these was misclassified. The algorithm also seems to perform fairly well for pure-breed dogs.On the other hand, there's clearly room for improvement, particularly around mutts. A little over half of them tend to be classified as a purebred (ie. a high degree of confidence), which is incorrect. However, this is not surprising, given the network was fully trained on purebreds and the fact that it recognizes close to half of these cases is nothing short of a great example of the power of CNNs.
As for points of improvement, the first thought revolves around the size of the training data. 6680 training samples seems like a good dataset until we consider than there are 133 breeds to classify (and, ideally, do a better job with mutts), which means we have roughly 50 samples per breed, with some as little as 26! That's really not much.
A second approach would be training the last few layers of Xception as well besides fine-tunning. This would require some experimentation, but given the datasets are not widely different (after all, ImageNet has dog breeds), it's very likely that additional improvements could be made there.
Lastly, augmenting the training data didn't yield any improvements when only fine-tunning, but if we were to train the last few layers of the pre-trained CNM, my intuition tells me it could have a more significant impact in that scenario.
One final improvement that could be made would be around efficiency. Given the nature of this jupyter notebook, there's a lot of redundancy and repeated work. The algorithm itself doesn't feel instant, which is understandable when we realize that, due to code reuse and abstractions, each image is loaded, parsed and treated multiple times for a single run! Having a much faster algorithm would provide a much better UX if this were to be used in a real-world scenario, like an app or service.
# Used for extracting bottleneck features. Initializing it is somewhat expensive and doing it for each function call
# is unecessary and wasteful. Hence, it's pre-initialized here.
bottleneck_model = Xception(weights='imagenet', include_top=False)
def Xception_predict_breeds(img_path):
from keras.applications.xception import preprocess_input
# extract bottleneck features
bottleneck_feature = bottleneck_model.predict(preprocess_input(path_to_tensor(img_path)))
# obtain predicted vector
predicted_vector = Xception_model.predict(bottleneck_feature)
# return dog breeds that are predicted by the model
first_index = np.argmax(predicted_vector)
first_probability = predicted_vector[0][first_index]
predicted_vector[0][first_index] = 0
second_index = np.argmax(predicted_vector)
second_probability = predicted_vector[0][second_index]
if (second_probability > first_probability / 3):
return [dog_names[first_index], dog_names[second_index]]
else:
return [dog_names[first_index]]
def render(plt, rows, columns, i, test_file, is_dog, is_human, breed_predictions):
plt.subplot(rows, columns, i + 1)
if is_dog:
plt.title("You are a dog!")
if (len(breed_predictions) == 1):
plt.xlabel("You are a " + breed_predictions[0])
else:
plt.xlabel("You are a mix of " + " and ".join(breed_predictions))
elif is_human:
plt.title("You are a human!")
if (len(breed_predictions) == 1):
plt.xlabel("You resemble a " + breed_predictions[0])
else:
plt.xlabel("You resemble a mix of " + " and ".join(breed_predictions))
else:
plt.title("I don't know what you are :(")
plt.xticks([])
plt.yticks([])
plt.imshow(cv2.cvtColor(cv2.imread(test_file), cv2.COLOR_BGR2RGB))
test_files = np.array(sorted(glob("testimages/*/*")))
count = len(test_files)
plt.figure(figsize=(20, count * 3))
columns = 2
rows = count / columns
for i, test_file in enumerate(test_files):
if (dog_detector(test_file)):
render(plt, rows, columns, i, test_file, True, False, Xception_predict_breeds(test_file))
elif (face_detector3(test_file)):
render(plt, rows, columns, i, test_file, False, True, Xception_predict_breeds(test_file))
else:
render(plt, rows, columns, i, test_file, False, False, [])
plt.show()
